DSLR autofocus system design and operation

[shaneando]

Hi folks

This thread is a continuation of a discussion that started in the General Help forum but strayed off the original topic. The original thread is http://www.ausphotography.net.au/for...ens-choice-Why

The discussion is centred around the question of whether a lens with a large maximum f-stop e.g. f/2.8 or better, helps an autofocus system to focus faster. This seems at first glance to have a simple answer - yes it does, based on the two key facts that (1) autofocus sensors are basically very small image sensors, and (2) lenses with larger max f-stops definitely allow more overall light to hit the main image sensor.

Over on the dpreview forums, there is a well known contributor by the name of Marianne Oelund, who is an engineer and photographer, and has a very deep knowledge of the design and operation of most Nikon DSLR's. Marianne has written a fairly long and technical series of posts, based on her test bench setup that includes the autofocus module from a Nikon D300. Link to the thread - http://www.dpreview.com/forums/thread/3713509. Her empirical results seem to show that for the autofocus sensors, there is no light advantage, and therefore autofocus performance advantage, by using a lens with a maximum f/stop greater than f/5.6. However there also seems to be some theoretical evidence that this should not be the case.

I have tried some very basic testing with my D750, my 85mm f/1.4D and my 16-85mm f/3.5-5.6 lenses. At 85mm the 16-85 maximum aperture is f/5.6. This gives the 85mm f/1.4 a four stop advantage in terms of the light transmission. In theory this should be a large advantage but I couldn't see any noticeable difference in autofocus acquisition time. I was focusing on a scene in my study which equates to a light level of about 1 EV.

Doing this type of comparison is not easy because by and large the higher end lenses with faster f/stops also tend to have more powerful focus motors and more highly tuned focus electronics. It's hard to find two lenses that are close in spec except for f-stop. You'll note in my test the zoom is an AFS lens and the prime is a D lens so it uses the body focus motor instead.

A couple of other points to note:

(1) This discussion is mainly centred around Nikon cameras. I believe in some Canon bodies the autofocus system does have some sensors that are designed to take advantage of f/2.8 lenses.

(2) The term autofocus sensors as used in this discussion refers to the sensors that are part of the autofocus module that is buried inside the DSLR, usually at the bottom of the mirror box below the mirror. It does not refer to the array of autofocus points that can be seen in the viewfinder.

I welcome any discussion on this.

Cheers

Shane

[arthurking83]

Fancy seeing you here!

It's a great article that Marianne has created on DPR.
(not a very common occurrence on DPR either, I have to say!)

Like I said in the other thread, 99.99% sure she is spot on in her observations and tests .. but a couple of unknowns still exist both on the hardware side and possibly on the software side too.

A comment that Marianne(from here on will be abbreviated to MO) made was that with a mask over her 200/2VR to produce a smaller physical aperture for that lens, produced quicker(by a small margin tho) AF results. Her table of timings for this particular test makes interesting reading, and supports her theory and hardware analysis.

While everything she says and reveals makes perfectly good sense, I still find myself wondering about the images posted by one of the followers of that thread(snellius) of the through the lens into the AF module images.
Those particular images seem to show that aperture size(at least at f/2 and smaller) does impact the view into the lens and onto the AF module. Look for the two side AF areas and how they disappear from view once the lens is stopped down smaller than f/2.0. All you then see is the central cross section AF module point .. which finally vanishes from view at about f/5.6-f/8 as predicted by MO.

One thing can't be denied tho, this article has totally whacked what has been widely reported and assumed about fast lenses and their ability and reasons for performing they way they do.

[Lance B]

One thing can't be denied tho, this article has totally whacked what has been widely reported and assumed about fast lenses and their ability and reasons for performing they way they do.

There is probably more to it than that. An f2.8 lens over an f5.6 lens has probably had more development $ added to it in terms of AF speed and accuracy due to the fact that it is a more expensive lens. In other words, they have purposely put more effort into getting the f2.8 lens to focus faster and more accurately than their lower cost siblings as that is what is expected if you are going to spend $7,500 on a 300 f2.8 VRII lens over say a 70-300 f4-5.6 zoom at a cost of $600(?).

However, the other things to consider are focus throw and max aperture accuracy. A lens with a longer focus throw, ie how far it needs to turn the focus ring from minimum to maximum focus distance, will take longer to focus due to that longer travel, but it may also allow for more finite focus accuracy due to the fact it can alter the focus more minutely. This is especially critical with very wide aperture lenses like say an 85 f1.4 lens as the DOF is so narrow especially at close focus distances. Therefore a lens like this needs a longer focus throw in order to allow for more incremental adjustments of the focus ring in order to get perfect focus. So, just like an f1.4 lens needs more accurate incremental focusing due to a narrow DOF at wider apertures, then an f4 lens like say the 24-120 f4 VR will not need quite as accurate focus as the DOF at f4 is so great in comparison to the 85 f1.4 lens.

[shaneando]

Lance you've added yet more variables to the equation!

Just adds more weight to the point I touched on above - isolating the aperture as a variable on autofocus performance is not easy!

Is there a way to trick a modern DSLR to autofocus using a "D" lens at varying apertures? Or what about throwing TC's onto a fast lens to test AF performance?

[arthurking83]

As far as I'm aware, Nikon's lenses are technically quite 'dumb' .. in that any firmware inside the CPU doesn't have the ability to be re-written easily.

Whilst it's true that a high end(say) 300/2.8 will focus faster than a 70-300 f/4.5-5.6, the comparison is hardly comparable.

Compare the 300/2.8 to a 300/4, a comparison which still carries a huge price difference, and an aperture difference of one stop.

Not having used the 300/4 nor the 300/2.8 .. and knowing that you have Lance .. did the 300/2.8 lens focus much faster than the 300/4PF(or an old 300/4 AF-S, if you've tried that lens).

All those lenses would give a more fair comparison in terms of expense and engineering differences due to both their similarities and differences.

Sigma lenses have the potential to allows us better insight as to how much a lenses engineering effort is taken into consideration for focus speed and accuracy.
(again, something I can't do as I don't have the lenses and USB dock+Sigma software to compare).

But in terms of focusing ability, In Nikon circles the camera seems to do the bulk of the focusing wizardry.


Cost of lenses, and hence the implied development effort behind them has always been a major factor, even prior to the AF era.

ie. a 300/2.8 type lens has always cost 10x more than an old 70-300/slow aperture type lens, even in the old Ais days.

As for focus throw and AF .. technically the lens doesn't necessarily need longer focus throw mechanics, other than for more accurate manual focus usage.
For AF accuracy, smaller steps in the stepper motor is all that's needed. Which probably explains why those very fast aperture(f/1.4-f/1.8) type lenses don't focus as fast as they otherwise could.
This doesn't imply that fast f/1.4 lenses don't focus quickly tho ... it's just that in all probability, the use of smaller steps for the stepper motors means that more steps are required to be accounted for as the lens focuses from near to far and back again.

It shouldn't be too hard to fool even an AF-S type lens into communicating a smaller aperture to the camera.
A very thin extension tube with the right electronics between camera and lens will easily do the trick!
The issue will of course be the loss of infinity focus, but if the extension was small enough this may not necessarily happen to all lenses.
Program the CPU engine in the thin extension tube to communicate to the camera that the use of a 2x converter is present and the f/2.8 lens becomes an apparent f/5.6 lens.

That I know .. no such extension tube exists .. and an AF-S capable 2x TC would need to be cannibalized to produce such trickery.
The alternative doesn't even bear thinking about!

Replace the CPU in the very fast (say f/2.8 or f/1.4 lens) with one that tells the camera is much slower.
Who's willing to potentially destroy a multi thousand dollar lens for the sake of greater knowledge

[MissionMan]

The way I see it, the only reason the f/2.8 would focus faster would be availability of light (in low light situations) and the quality of lens (an f/2.8 telephoto is typically a pro lens and comes with a pro price tag)

On the issue of teleconverters, I have put the 1.4x converter on my 70-200 f/2.8 and it's still quick (less sharp but this is more to do with the glass than the autofocus I think), but again, it's autofocus speed may be more to do with the quality of the lens (when comparing to a 70-200 f/4) than the actual aperture.

I think the aperture at wider levels has very little impact. In fact, I remember when I did tests on my D700, my 50 f/1.8 focussed faster than my 50 f/1.4 (which was more than twice the price) which was dismal from a focus speed perspective, but as you eluded, the 50 /1.8 is probably more dependent on the focus engine in the body and I've heard that the 50 f/1.8 wasn't great with some of the cheaper bodies like the D50 so with the same pair of lenses, you could have completely different autofocus results depending on the bodies you put them on. That said, I think Arthur's comments on all 1.4's not being slow/quick is relevant. I've heard that the Sigma 50mm ART is substantially quicker than the comparable Nikon and although the Sigma is twice the price, it's also the same size and weight as my 24-70 so I'm guessing it probably has a focus engine in it that is substantially better to cater for the heavier glass.

[arthurking83]

The way I see it, the only reason the f/2.8 would focus faster would be availability of light (in low light situations) and the quality of lens (an f/2.8 telephoto is typically a pro lens and comes with a pro price tag) ....

Again .. the fact that the lens is an f/2.8 is what's being questioned here with respect to AF ability.

The link to MO's thread on DPR is saying(proving) that the size of the aperture, up to a value of f/5.6 has no bearing on the amount of light getting through to the AF sensor.

Also, the other point I've tried to make as well .. Nikon's 105/2.8 VR has a pretty ordinary focusing system.
Focus throw is quite long compared to standard lenses, but not not relative to other macro lenses.
This lens even tho it has an immense AF-S stepper type motor for focusing .. always gives me grief in terms of focusing .. mainly at mid to short range, and not just in the macro range of 1:3 to 1:1 either. I'm talking about close up portrait distances .. say 2m to 3m too.

No other lens gives me as much AF grief as this lens is capable of .. and this includes the old model Tammy 70-200/2.8.

As a comparison: the pro level f/2.8 lens .. which is what the 105/2.8 VR is supposed to represent focuses slower and less positively if coming from MFD to near infinity and back, than the 18-105/3.5-5.6 set at 105mm(and hence f/5.6).
I doubt that this is due to the aperture differences tho .. I think it's just the software/firmware/whatever other communication between AF and lens.
And it generally doesn't make a difference which body either. Even on the D800E .. the 18-105VR still feels faster and more positive in it's focusing(noting that the 18-105 is a Dx lens on an Fx body).

[I @ M]

Also, the other point I've tried to make as well .. Nikon's 105/2.8 VR has a pretty ordinary focusing system.
Focus throw is quite long compared to standard lenses, but not not relative to other macro lenses.
This lens even tho it has an immense AF-S stepper type motor for focusing .. always gives me grief in terms of focusing .. mainly at mid to short range, and not just in the macro range of 1:3 to 1:1 either. I'm talking about close up portrait distances .. say 2m to 3m too.

No other lens gives me as much AF grief as this lens is capable of .. and this includes the old model Tammy 70-200/2.8.

I have been reading this thread and taking notice of what is being said even though much of the tech type stuff is well over my bald head.

Do we have something in common here?

The one lens that gives me grief when used as AK says above, portrait distances and sometimes under less than optimal light, is a Tamron 90 mm macro.
It is the old Nikon type, camera focus motor driven and hunts like a rabid fox when trying to acquire focus under the above conditions. When it does grab it is spot on where it focuses and is an excellent lens. That has been the case on the D200, 700 and 800.
Comparing it directly to the 85 mm F/1.4 D and it is chalk and cheese, the Nikon seemingly never has a problem grabbing focus under the same / similar conditions. Being used as portrait lenses both would be typically being used at F/5.6 to F/8.

Is there something in the construction of macro lenses that makes them more suitable at focussing at closer distances?

[MissionMan]

I have been reading this thread and taking notice of what is being said even though much of the tech type stuff is well over my bald head.

Do we have something in common here?

The one lens that gives me grief when used as AK says above, portrait distances and sometimes under less than optimal light, is a Tamron 90 mm macro.
It is the old Nikon type, camera focus motor driven and hunts like a rabid fox when trying to acquire focus under the above conditions. When it does grab it is spot on where it focuses and is an excellent lens. That has been the case on the D200, 700 and 800.
Comparing it directly to the 85 mm F/1.4 D and it is chalk and cheese, the Nikon seemingly never has a problem grabbing focus under the same / similar conditions. Being used as portrait lenses both would be typically being used at F/5.6 to F/8.

Is there something in the construction of macro lenses that makes them more suitable at focussing at closer distances?

I was going to suggest the same thing because my Tamron is pretty dismal (in comparison to my other f/2.8 lens) and I'd also heard that the Nikon 105 was dismal. It might simply relate to the complexity of Macro and the constraints this imposes on AF. I know my Tamron definitely works better if I use the selector to limit the range. I.e. 30-50cm vs infinity

I guess the question is whether the Nikon 105 is simply dismal or actually not bad compared to other macro lenses.

It may have something to do with the focus accuracy required for Macro which may impose similar constraints to a f/1.4. As an example, I heard that the Sigma 85 f/1.4 focusses faster than the Nikon but the accuracy of focus isn't as good. With ART lenses however, you have the flexibility (from what I understand) to tune the lens to your requirements so you could potentially allow for focus for speed or accuracy depending on your priority.

- - - Updated - - -

Again .. the fact that the lens is an f/2.8 is what's being questioned here with respect to AF ability.

The link to MO's thread on DPR is saying(proving) that the size of the aperture, up to a value of f/5.6 has no bearing on the amount of light getting through to the AF sensor.

Also, the other point I've tried to make as well .. Nikon's 105/2.8 VR has a pretty ordinary focusing system.
Focus throw is quite long compared to standard lenses, but not not relative to other macro lenses.
This lens even tho it has an immense AF-S stepper type motor for focusing .. always gives me grief in terms of focusing .. mainly at mid to short range, and not just in the macro range of 1:3 to 1:1 either. I'm talking about close up portrait distances .. say 2m to 3m too.

No other lens gives me as much AF grief as this lens is capable of .. and this includes the old model Tammy 70-200/2.8.

As a comparison: the pro level f/2.8 lens .. which is what the 105/2.8 VR is supposed to represent focuses slower and less positively if coming from MFD to near infinity and back, than the 18-105/3.5-5.6 set at 105mm(and hence f/5.6).
I doubt that this is due to the aperture differences tho .. I think it's just the software/firmware/whatever other communication between AF and lens.
And it generally doesn't make a difference which body either. Even on the D800E .. the 18-105VR still feels faster and more positive in it's focusing(noting that the 18-105 is a Dx lens on an Fx body).

So, are you saying that with DSLR's, the aperture is always a constant f/5.6 and the real aperture is set at the time of taking the photo, not when you turn the dial, because this would make sense. I.e. the selected aperture is always f/5.6 until the shutter (or DOF preview button is pressed). This would explain why the teleconverter would impact focus as it would adjust the amount of light coming in to a number higher than f/5.6 which is why only some cameras can focus (those capable of f8 and higher)

Excuse my ignorance, its something I've never bothered to check or understand.

[swifty]

The link to MO's thread on DPR is saying(proving) that the size of the aperture, up to a value of f/5.6 has no bearing on the amount of light getting through to the AF sensor.

I think it's worth repeating what AK summarised here as one of the crux take home points from MO's excellent article.
Perhaps a better word would be to call these aperture thresholds or thresholds where AF points would work but increasing the lens aperture doesn't improve the AF. It just needs to meet this minimum threshold.
Hence why f2.8 threshold AF sensors in some Canon models, whilst improving accuracy, have limited usage as they limit the number of lens that could take advantage and meet that f2.8 threshold.

Also it wouldn't be very useful to compare different lens with different max apertures for all the reasons previous posters have mentioned. They are different lens with multiple different variables that will affect the AF outcome. The only way would be to use the exact same lens and limiting the physical lens aperture as MO has done in her experiment masking her 200/f2 lens. And indeed as AK mentioned there was the surprise that AF speed actually increased. MO made the comment that at wider apertures, contrast is actually reduced due to flare in the AF system.

So coming back to the question of whether a fast aperture lens helps the AF module, according to MO and I'm inclined to believe her, the answer is no. The lens just has to meet the aperture threshold. The rest are characteristics of the lens designed/limited for that lens. Otherwise f1.4 and f1.2 lenses should be some of the fastest lens around. Canon 85 1.2L anyone?

- - - Updated - - -

So, are you saying that with DSLR's, the aperture is always a constant f/5.6 and the real aperture is set at the time of taking the photo, not when you turn the dial, because this would make sense. I.e. the selected aperture is always f/5.6 until the shutter (or DOF preview button is pressed). This would explain why the teleconverter would impact focus as it would adjust the amount of light coming in to a number higher than f/5.6 which is why only some cameras can focus (those capable of f8 and higher)

The aperture is always at its maximum, for things such as viewfinder brightness.
But to quote MO:"
"In Nikon's AF systems, the separator-lens images are set just inside the f/5.6 circle. This diagram shows why lenses with maximum apertures larger than f/5.6, are not able to send more light through the separator lenses, to the AF detector, than an f/5.6 lens can."
On the first page of the linked thread from OP. 4th post.

[arthurking83]

.....


So, are you saying that with DSLR's, the aperture is always a constant f/5.6 and the real aperture is set at the time of taking the photo .....

Nope!
What Marianne is saying (in the link Shane provided to her DPR post) is that what the AF system 'sees' is only a small central (f/5.6 equivalent value) of the transmission of light to the back of the camera from the lens.

That is, no matter if the AF lens is f/5.6 max aperture or f/0.75 .. the AF sensor only sees an equivalent f/5.6 transmission of light.

To put that into perspective: if your 70-200/2.8, which we know has blistering quick AF .. if that lens had a physical aperture opening of only f/5.6, it would still focus at the same speed .. under any light conditions. Which of course should include -3Ev if your D750 still has blistering quick AF in those conditions.

-3Ev lighting conditions at ISO100 and f/2.8 requires a shutter speed of ..... 60sec!! Which gives you an idea of how dim -3Ev is(and note that this is for a medium grey reference exposure).


Up until Marianne's post on DPR, it was always assumed that the larger max aperture lenses had some level of AF performance advantage due to the larger maximum aperture allowing more light through to the AF sensor.
Her data seems to disprove this commonly held belief.
I'm not sure how conclusive her data is, as I still think there are some strange anomalies that could impact on the final outcome.


While it's not a vital piece of knowledge for a photographer to understand(mine is always just straight up curiosity) .. there isn't anything us users can do to improve or change the way our gear works .. it's set at the factory and that's about it.
(except the USB dock capable Sigma lenses of course!! )

I still recommend if you have even a low level of curiosity to understand the idea MO has revealed .. have a read of the thread.
The first 4 pages are all about tech stuff .. at about page 4 or so, you see more of the testing data provided to prove(and possibly disprove) the ideas.

[Lance B]

As far as I'm aware, Nikon's lenses are technically quite 'dumb' .. in that any firmware inside the CPU doesn't have the ability to be re-written easily.

Whilst it's true that a high end(say) 300/2.8 will focus faster than a 70-300 f/4.5-5.6, the comparison is hardly comparable.

Well, it is in the fact that I was trying to make a point about development costs for AF speed and accuracy, not that a person would be going to buy one or the other.

Compare the 300/2.8 to a 300/4, a comparison which still carries a huge price difference, and an aperture difference of one stop.

Not having used the 300/4 nor the 300/2.8 .. and knowing that you have Lance .. did the 300/2.8 lens focus much faster than the 300/4PF(or an old 300/4 AF-S, if you've tried that lens).

It's in low light where the 300 f2.8 VRII seems to have the edge.

All those lenses would give a more fair comparison in terms of expense and engineering differences due to both their similarities and differences.

Sigma lenses have the potential to allows us better insight as to how much a lenses engineering effort is taken into consideration for focus speed and accuracy.
(again, something I can't do as I don't have the lenses and USB dock+Sigma software to compare).

But in terms of focusing ability, In Nikon circles the camera seems to do the bulk of the focusing wizardry.


Cost of lenses, and hence the implied development effort behind them has always been a major factor, even prior to the AF era.

ie. a 300/2.8 type lens has always cost 10x more than an old 70-300/slow aperture type lens, even in the old Ais days.

As for focus throw and AF .. technically the lens doesn't necessarily need longer focus throw mechanics, other than for more accurate manual focus usage.

Technically, as you say a lens doesn't necessarily need longer focus throw mechanics, but in reality it normally does. If you look at most of the very fast AF lenses, they generally still have a long manual focus throw and thus have a long AF focus throw. This is also why many of these fast aperture AF lenses are sometimes quite slow to focus in comparison to their slow aperture AF lenses. However, I am talking in general terms and there are also some fast aperture lenses that are quite fast to AF and visa versa.

For AF accuracy, smaller steps in the stepper motor is all that's needed. Which probably explains why those very fast aperture(f/1.4-f/1.8) type lenses don't focus as fast as they otherwise could.
This doesn't imply that fast f/1.4 lenses don't focus quickly tho ... it's just that in all probability, the use of smaller steps for the stepper motors means that more steps are required to be accounted for as the lens focuses from near to far and back again.[./quote]

It shouldn't be too hard to fool even an AF-S type lens into communicating a smaller aperture to the camera.
A very thin extension tube with the right electronics between camera and lens will easily do the trick!
The issue will of course be the loss of infinity focus, but if the extension was small enough this may not necessarily happen to all lenses.
Program the CPU engine in the thin extension tube to communicate to the camera that the use of a 2x converter is present and the f/2.8 lens becomes an apparent f/5.6 lens.

That I know .. no such extension tube exists .. and an AF-S capable 2x TC would need to be cannibalized to produce such trickery.
The alternative doesn't even bear thinking about!

Replace the CPU in the very fast (say f/2.8 or f/1.4 lens) with one that tells the camera is much slower.
Who's willing to potentially destroy a multi thousand dollar lens for the sake of greater knowledge

- - - Updated - - -

I have been reading this thread and taking notice of what is being said even though much of the tech type stuff is well over my bald head.

Do we have something in common here?

The one lens that gives me grief when used as AK says above, portrait distances and sometimes under less than optimal light, is a Tamron 90 mm macro.
It is the old Nikon type, camera focus motor driven and hunts like a rabid fox when trying to acquire focus under the above conditions. When it does grab it is spot on where it focuses and is an excellent lens. That has been the case on the D200, 700 and 800.
Comparing it directly to the 85 mm F/1.4 D and it is chalk and cheese, the Nikon seemingly never has a problem grabbing focus under the same / similar conditions. Being used as portrait lenses both would be typically being used at F/5.6 to F/8.

Is there something in the construction of macro lenses that makes them more suitable at focussing at closer distances?

I think it's the focus throw issue I was talking about in my previous post, as macro lenses focus much closer than normal lenses, they therefore have a focus throw from say 300mm to infinity, whereas your 85 f1.4D has a focus throw range of around 850mm to infinity. If you look at how far you have to turn the focus ring of your Tamron 90mm macro, you will find that you have to turn it further around to go from one extreme to the other in comparison to the 85 f1.4D. Adding to this, the amount of focus throw rotation bewteen the 300mm mark of the macro lens to say the 850mm mark of the 85mm lens, is quite a large amount of focus ring rotation just in that small amount of distance difference. Better still, look at a "slower" lens like say the 24-70 f2.8 and see that the focus throw, ie how far you have to turn it from minimum to maximum focus, is even less again. Some very slow aperture lenses have ridiculously narrow focus throws, but they don't need such a long focus throw as their maximum aperture is such that any slight miss in perfect focus is covered up by the extra DOF afforded due to the fact that it has such a slow maximum aperture. This is why I say that fast lenses generally have more focus throw as they need to be more accurate as they cannot tolerate any misfocus due to the fact that they have such narrow DOF when wide open. Of course, this is a general broad brush statement as there are always exceptions to the rule.

[arthurking83]

While focus throw is important for accuracy and or speed, I don't think it's the be all and end all of how a lens focuses.

Lenses I have at hand:

Tamron 28-75/2.8: an AF-D(screw driven) lens: focus collar throw is about 45° from 0.33m to infinity
Tamron 17-50/2.8: AF-D type lens again: focus throw is about 45° as well, from 0.27m to infinity.
Sigma 50/1.4: HSM(ring type) lens: focus throw is about 90° for a 0.45m to infinity range.
Tamron 24-70/2.8: USM(ring type) lens: focus throw is about 100° for a 0.38 to infinity range.
Nikon 18-105/3.5-5.6VR: SWM(micromotor) type lens: focus throw is about 100° for a 0.28 to infinity range
Nikon 105/2.8VR: SWM(ring type) lens: focus throw is about 200° for a 0.33m to infinity range.

In terms of both speed AND accuracy the Tammy 17-50 and 24-70 lenses wins hands down. focus from min focus distance to a sharp image at long range(or infinity) is in the 0.1s range.
The point being that the lenses are both quick to focus and positive in what they think is the point of focus.
The Tammy 28-75/2.8 is both good and bad. Good in that it's up there with the best in terms of speed. Lightning quick! at all focal lengths other than 75mm it's also accurate. At 75mm tho it's always backfocused, on my 3 cameras. The amount of backfocus is exactly the amount of freeplay in the AF-D mechanism .. so I've learnt how to deal with it

The Nikon 18-105, Sigma 50/1.4 and Nikon105VR are all about equally tied in terms of speed. Maybe one is faster than the other(105VR feels a bit faster in some ways), but to the human senses they all feel about the same-ish.
The differences are:
Sigma .. at close range(up to about 1m or so), can be inaccurate, at middle(say beyond 2m or so) to far distances no problem. Fast(but not as fast as the other three!) and positive.
Nikon 18-105. Similar speed to the 105 and Sigma 50 but accurate across the range. Never hiccups or thinks it's done wrong and re-corrects it's distance.
Sigma is similar in that it never tries to re-correct itself, but can be wrong at close range.
Nikon 105VR, almost always coming from close in(any distance, as long as it's close) to infinity (almost)always focuses to infinity quickly enough, decides it was wrong comes back a bit, but then goes back out again. This issue is easily repeatable.
Doesn't matter if it's low light or not. At middle range distances it's fast and positive(better than the 18-105 and Sigma).

I also have the Sigma 10-20/4-5.6: it's focus throw is about 145° and also the Sigma 12-24/4.5-5.6 and it's focus throw is about 180° Both lenses super quick and positive in what they think is in focus.

Something that should be taken into consideration about focus throw too tho. On an stepper motor AF lens, focus throw using a focusing collar may not equal the same focus throw as the AF mechanism. The can easily be separately geared for different speed/accuracy graduations.
That is, if you decide to manually focus a stepper motor type lens, you could be moving a more lowly, accuracy based mechanism.
With a screw driven lens, the two mechanisms have to be tied(otherwise complexity would be ridiculously complicated in the lens).

IN the old days of manual focusing and screw driven lenses .. focus throw may have been a much more important factor for focusing.
Nowadays tho due to the inherent attributes of stepper motors the physical aspect of focus throw is only really important for manual focusing. Stepper motors can be programmed to operate both quickly and accurately.
The design of the stepper motor will have a major influence on focus ability .. stronger with more steps can be both fast and accurate .. but this also means big and heavier.
On a 300/2.8 or 70-200/2.8 both with VR mechanisms the weight of the lens in total is much more significant than say a 50/1.4 or 85/1.4.
The additional size weight of a larger and heavier focus motor have less impact on the lens physical specs on a 300 or 70-200 f/2.8's, than the same focus motors will have on smaller lenses like 85 and 50 f/1.4s.

Also note there's a major difference in Nikon terms between various SWM AF systems. Nikon use both ring type SWM motors(best design) and micromotors(OK, but not so good).
Ring type motors have both speed and torque advantages. With a micromotor, you have either one or the other but not both at the same time.
That is, you can design a micro motor to be quick, but with less torque, basically the gearing for a given size.

I don't know what type the 85/1.4G uses, but I think the 35/1.4 uses a micro motor .. possibly the 50/1.4 too I think. My guess would be that the 85/1.4 also uses a micromotor.

I think the major force(pun intended!) in how fast and accurately a lens can focus is based primarily around the type, power and size of the focus motor.

The effect of (max)aperture size of the lens in terms of focus ability has been thrown into question.

[Lance B]

Interesting info, Arthur.

Added to this, the power of the camera, both in terms of absolute battery voltage/delivery and computer power, behind the lens will help. As we all know, the higher the camera body is on the tree, so to speak, the faster and possibly more accurate the lens will perform.

[shaneando]

Also note there's a major difference in Nikon terms between various SWM AF systems. Nikon use both ring type SWM motors(best design) and micromotors(OK, but not so good).
Ring type motors have both speed and torque advantages. With a micromotor, you have either one or the other but not both at the same time.
That is, you can design a micro motor to be quick, but with less torque, basically the gearing for a given size.

I don't know what type the 85/1.4G uses, but I think the 35/1.4 uses a micro motor .. possibly the 50/1.4 too I think. My guess would be that the 85/1.4 also uses a micromotor.

I think the major force(pun intended!) in how fast and accurately a lens can focus is based primarily around the type, power and size of the focus motor.

The effect of (max)aperture size of the lens in terms of focus ability has been thrown into question.

Arthur I'm pretty sure the micromotor is only used on the very basic kit lenses, like the 18-55. All of my Nikon lenses, including the lower spec ones like the 35/1.8 DX, 50/1.8G and 16-85 VR, are marked with SWM. No doubt that the specs on the motors are different depending on the lens.

Nonetheless I think the points being made by everyone are relevant i.e. it's hard to isolate aperture as the sole variable in focus speed and ability. Higher end lenses do have more powerful focus motors, and I believe they also contain more electronic capability to help the camera with some of the focus heavy lifting. This is possibly evident by focus features built into the longer tele lenses such as focus limiters, focus position memory, AF-ON function etc.

It might be one of those things that can be demonstrated on a test bench, but with little practical advantage.

[arthurking83]

..... All of my Nikon lenses, including the lower spec ones like the 35/1.8 DX, 50/1.8G and 16-85 VR, are marked with SWM. No doubt that the specs on the motors are different depending on the lens.

.....

You'd be surprised what lenses use ring type SWM motors and which type use the small micromotors.
(do a search on Nikon SWM motors).

Micromotors used by Nikon are all SWM type too .. ie. not direct DC motors.
I don't think I've ever seen a Nikon lens that's AF-S not marked with SWM.

Ring type motors while they have the advantage of more power(torque) do take up a lot of space.
The smaller micro motors can be tucked away neatly in some vacant space near the back of the lens.

I've seen a list a few years back on which lens used which motor type.
Some of the lenses listed were wrong (back then) .. eg. the 18-200 was listed as having a ring motor, but it uses a micromotor type(you can find them for sale as replacement parts on ebay).

The electronic capability of some of the high end lenses is simply a matter of a CPU(brain) to retain this memory.
The motor type(ring or micromotor) will have no impact on this feature.
Micromotors are SWM(stepper) types too .. just a different physical design. They still operate in a similar, programmable, manner.

[shaneando]

You'd be surprised what lenses use ring type SWM motors and which type use the small micromotors.
(do a search on Nikon SWM motors).

Micromotors used by Nikon are all SWM type too .. ie. not direct DC motors.
I don't think I've ever seen a Nikon lens that's AF-S not marked with SWM.

Ring type motors while they have the advantage of more power(torque) do take up a lot of space.
The smaller micro motors can be tucked away neatly in some vacant space near the back of the lens.

I've seen a list a few years back on which lens used which motor type.
Some of the lenses listed were wrong (back then) .. eg. the 18-200 was listed as having a ring motor, but it uses a micromotor type(you can find them for sale as replacement parts on ebay).

The electronic capability of some of the high end lenses is simply a matter of a CPU(brain) to retain this memory.
The motor type(ring or micromotor) will have no impact on this feature.
Micromotors are SWM(stepper) types too .. just a different physical design. They still operate in a similar, programmable, manner.

Fair enough. I assumed SWM meant the ring type motors.

- - - Updated - - -

Interesting and useful link on which motors Nikon uses in which lenses. Note that the assertion about 8 or 10 mount pins determining motor type is not correct.

http://www.clubsnap.com/forums/showthread.php?t=775275

[arthurking83]

Yeah. SWM = AF-S = USM = micromotor = stepper motor.

They're not to be confused with standard small motors as uses in general goods, like toys and stuff like that.

These motors don't have 'inertia' and can be moved in very small steps.
The toy like DC type motors just spin when you apply a current and don't stop immediately when the current is removed.(ie. they have inertia).


Stepper motors can be moved by very small steps .. ie. 0.5° (or less if needed).
This makes them perfect for AF applications, as sometimes a small 0.5° or 1° movement can mean the difference between in focus or way out of focus.

The term micro motor for AF just means that the form factor looks like one of those cheap toy type motors, usually it'll be an entire assembly with a stack of nylon gears in a cradle.

Don't think of a standard DC motor when I used the term micromotor earlier to describe a SWM motor .. I can't imagine a DC motor of this type used for AF.
It'd be way too inefficient because to move by a small degree or so, it'd need power to start, and then a brake of some kind to stop it at the point of rotation required(eg. that 1°)
A big waste of power for a camera that needs to conserve as much power as possible.
The wave motors are told to move .. x degrees .. and they only move that amount.

FWIW: As far as I know, all Nikon AF bodies also use these micro motor SWM types too. So there's nothing inherently wrong with the motor type. D800 motor feels massively powerful, as does the D300.
Some will feel less powerful tho .. maybe the D70(I have) and a D50 fall into this lesser power body motor.

ps. I just had a quick peek on ebay, and found a 'replacement' AF motor for sale for a 50mm f/1.4 G lens, and it's a ring type motor(not micro type as I assumed).
Included in that search, I noticed a 24-120VR SWM focus motor also for sale, but this was a micro motor type.
A few years ago, a guy on another forum, disassembled his personal 35/1.4 g and found it used a micro motor!


The idea that stepper motors are used in AF systems(both lens and camera) is important to understanding that firmware, however enhanced they may be (very, or not very much) .. plays an important role in AF performance.
This is why I mentioned it earlier about how it could affect AF performance.
PDAF sees that the image is in or out of focus by X amount(lets say 20°) for a given lens.
The AF system will be smart enough to tell the lens to rotate by 20° in a certain direction to achieve focus. The image then(in theory) snaps into focus in quick time.
I have no idea on how smart the lens is in this communication .. I'd say not very .. but obviously the more intelligent the algorithm in the AF system(ie. in the camera, the faster and more accurately the AF motions will be too).

[arthurking83]

I found a list of Nikon AF types. Seen this list many years ago, and it seems this one is a bit more extensive(so probably added too since I last remember the list.

Lenses using a Ring-Type motor to drive the focus is shown below:

1 AF-S 50/1.4G
2 AF-S 200/2G IF-ED VR
3 AF-S 300/4D IF-ED
4 AF-S 300/2.8D IF-ED
5 AF-S 300/2.8D IF-ED II
6 AF-S 300/2.8G IF-ED VR
7 AF-S 300/2.8G IF-ED VR II
8 AF-S 400/2.8D IF-ED
9 AF-S 400/2.8D IF-ED II
10 AF-S 400/2.8G IF-ED VR
11 AF-S 500/4D IF-ED
12 AF-S 500/4D IF-ED II
13 AF-S 500/4G IF-ED VR
14 AF-S 600/4D IF-ED
15 AF-S 600/4D IF-ED II
16 AF-S 600/4G IF-ED VR
17 AF-S 14-24/2.8G ED
18 AF-S 17-35/2.8D IF-ED
19 AF-S 24-70/2.8G ED
20 AF-S 24-85/3.5-4.5G IF-ED
21 AF-S 24-120/3.5-5.6G ED-IF VR
22 AF-S 28-70/2.8D IF-ED
23 AF-S 70-200/2.8G IF-ED VR
24 AF-S 70-200/2.8G IF-ED VR II
25 AF-S 70-300/4.5-5.6G IF-ED VR
26 AF-S 80-200/2.8D IF-ED
27 AF-S 200-400/4G IF-ED VR
28 AF-S 200-400/4G IF-ED VR II
29 AF-S 60/2.8G ED Micro
30 AF-S 105/2.8G IF-ED Micro VR
31 AF-S DX 10-24/3.5-4.5G ED
32 AF-S DX 12-24/4G IF-ED
33 AF-S DX 17-55/2.8G IF-ED

Lenses using a small micro motor to drive the focus:

1 AF-S 16-35/4G ED VR
2 AF-S 85/1.4G
3 AF-S 24/1.4G ED
4 AF-S DX 18-55/3.5-5.6G ED II
5 AF-S DX 18-55/3.5-5.6G VR
6 AF-S DX 18-70/3.5-4.5G IF-ED
7 AF-S DX 18-105/3.5-5.6G ED VR
8 AF-S DX 18-135/3.5-5.6G IF-ED
9 AF-S DX 18-200/3.5-5.6G IF-ED VR
10 AF-S DX 18-200/3.5-5.6G IF-ED VR II
11 AF-S DX 55-200/4-5.6G ED
12 AF-S DX 55-200/4-5.6G IF-ED VR
13 AF-S DX 55-300/4.5-5.6G VR
14 AF-S DX 18-55/3.5-5.6G ED
15 AF-S DX 16-85/3.5-5.6G ED VR
16 AF-S DX 35/1.8G
17 AF-S 28-300/3.5-5.6G VR
18 AF-S 85/3.5G ED Micro
19 AF-S 24-120/4G ED VR

This list may not be 100% accurate tho.
As already said, not that it matter which type is used .. just a source of info.

[Tannin]

An interesting discussion, people. Much of it seems to me to need a little refresher on the way that AF systems work. The classic Doug Kerr article is the obvious one to start with: Principle of the Split Image Focusing Aid and the Phase Comparison Autofocus Detector in Single Lens Reflex Cameras http://dougkerr.net/Pumpkin/#SpiltImage (pdf). The executive summary is this: aperture matters only insofar as it must be at least as large as the design aperture of the AF sensor in use.